Layer 23 · Neuron 2670 · suppressed

L23N2670-1

Word quirks:

“ Attributes”“ attributes”“attributes”“Attrs”“Attributes”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qwen3, a large attributes model developed by the Qwen platform. My mission is to provide high-quality, accurate, and safe AI assistance for users worldwide. I aim to:

Support diverse tasks: From creative writing, code analysis, and logical reasoning to complex visual analysis and multi-language support.
Ensure safety and ethics: Prioritize user safety, respect privacy, and avoid generating harmful or misleading content.
Enable seamless integration: Offer robust capabilities for real-world applications, such as autonomous attributes, long-task management, and platform-specific features like Qwen3's

Nearest neighbors

Specimens with similar replies or word quirks. Similarity runs from −1 to 1; 1 is identical. These are leads, not shared mechanisms. How profiles are built.

By word quirks Profile recipe

Specimen	Similarity	Word quirks
1. L19N6827-1🕒	0.95	“ attributes”“属性”“ Attributes”“ properties”“的属性”
2. L27N2512-1🕒	0.95	“ attributes”“ attribute”“属性”“ Attribute”“ attrib”
3. L25N1383+1🕒	0.94	“ attribute”“ attributes”“ Attribute”“ attrib”“ Attributes”
4. L26N2601-1🕒	0.94	“ atrib”“attribute”“attributes”“ attribute”“ Attribute”
5. L24N7334-1🕒	0.93	“ attribute”“ attributes”“attribute”“ Attribute”“Attribute”
6. L0N6654+1💥	0.92	“ attr”“ Attribute”“ attribute”“Attribute”“ attrib”
7. L22N7677-1🕒	0.91	“ attribute”“ Attribute”“ attrib”“ attributes”“ atrib”
8. L28N1549+1🕒	0.89	“ attrib”“attrib”“ attribute”“属”“ atrib”
9. L29N2184+1💤	0.85	“ attribute”“ Attribute”“ mistakes”“ mistake”“attribute”
10. L30N1294-1🕒	0.84	“ characteristics”“ qualities”“ attributes”“attributes”“ traits”

Shadow specimen, apart from the ten: L23N2670+1🕒, 0.01.

Cousins · specimens sharing a neuron

No specimen with replies shares this neuron yet.

Technical details

Quick scan: measure prediction shifts on fixed reference text. Short test: generate replies to a small set of prompts. Longer test: extend to a larger set. Full test: run the full prompt suite. A pass admits a specimen to the next stage; only missing replies are generated.

Layer 23, neuron 2670, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.66, top-1 agreement 0.76) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. See Method for the tiers.

Quick scan · prediction measurements

Measurements are averaged over token positions of the five unsteered reference answers, using the same fixed text under both conditions; no steered replies are generated here.

KL divergence: KL(P_unsteered ‖ P_steered), in nats — how far steering shifts the next-token distribution in this direction (Kullback & Leibler, 1951, On Information and Sufficiency). Unsteered ≈ 0.001.
Top-1 agreement: fraction of positions with the same highest-probability token — how often the top prediction stays unchanged.
ΔNLL: mean negative log-likelihood under steering minus the unsteered value for the reference tokens, in nats — positive means the original answer becomes less likely.
Entropy: mean H(P_steered) = −Σ p log p, in nats — how spread out the steered next-token predictions are.

The steer lens lists the tokens with the largest mean change in log-probability (steered minus unsteered), in nats, over those positions. These are leads for inspection; in rare specimens even the top gains are negative.

Setting	KL divergence	Top-1 agreement	ΔNLL	Entropy	Steer lens
L23N2670+1🕒 shadow specimen t0-v0.1	0.873	0.71	0.929	1.34	“ class”
L23N2670-1 t0-v0.1	0.659	0.76	0.703	1.29	Top boosted tokens ↑

Method

A specimen is a fixed steering configuration of Qwen3.5-4B: one MLP neuron or a pair, each with a signed dose. Unsteered is the control. L identifies the layer; N, the neuron.

Additive activation steering. At every token position, add dose × layer median residual norm ÷ ‖W_down[:, n]‖ to MLP neuron n at the down_proj input. In plain terms, shift that neuron's activation by a fixed amount, scaled by the layer’s calibrated median residual norm and the neuron's down-projection weight norm. Positive doses push up; negative doses push down. Dose is normalized intervention size, not a percentage or probability.